### Results Replication 

We release intermediate data that we use to produce Figures 3, 4, 8, 9, and 10, Table 7, and corresponding statistics in Section 5 (Results) and Sections 8 and 9 (Appendix).

1. Installation: `pip3 install -r requirements_results.txt`
2. Image analysis: `python3 img_analysis.py` and `python3 img_analysis.py --add_st_wt True`
3. Query analysis: `Rscript qry_analysis.R`

### Data Collection and Methods Scripts

Per our IRB protocol, we cannot release individual participant queries. Thus, we cannot make the individual-level data we relied on in Sections 3 (Data Collection) and 4 (Methods) public. However, for the sake of transparency, we do release the majority of the python scripts we used.

1. `https://github.com/jlgleason/google-image-scraper`: Google and Bing image scraper 
2. `person_detection.py`: YOLOv3 person detection on individual image files
3. `person_detection_sota.py`: CO-DETR person detection on individual image files
4. `spacy_ner.py`: identify named entities using Spacy's transformer/CNN pipelines
5. `google_knowledge_ner.py`: identify named entities using presence of Google Search special features
6. `nsfw_classifier.py`: identify NSFW images
7. `annotations.py`: MTCNN face detection on individual images for Mechanical Turk labeling
8. `annotations_sota.py`: SCRFD face detection on individual images for Mechanical Turk labeling